I am asked to extract three curves from palm images. enter image description here

               original image

enter image description here expected result. The curves has lower intensity than neighboring pixels. So I try to morphology to extract the pixels. The matlab code is followed.

RGBImage = imread('E:\ImageProcessing\Palm\Matlab\IMG_20140817_144549.jpg');
GrayImage = rgb2gray(RGBImage);
GrayImage = imresize(GrayImage, [768, 1024]);
se = strel('disk', 10);
GrayImage = imbothat(GrayImage, se);
figure; imshow(GrayImage);

The result is shown below enter image description here

The curves still have a little higher intensity values than neighboring pixels. Maybe I need use ridge filter to enhance the curves. I use the filter from matlab file exhange(http://www.mathworks.com/matlabcentral/fileexchange/24409-hessian-based-frangi-vesselness-filter). The filtered result is not satisfactory. I hope to get any advance or links or papers.

  • $\begingroup$ The palm image you have provided is blurry and illumination is not suitable for this type of detection. Do you have the liberty of choosing the conditions or are you provided with a palm images set ? $\endgroup$
    – Tyathalae
    Aug 20, 2014 at 6:39
  • $\begingroup$ Thanks. I am provided with one set of palm images. Can you clarify the reason that the illumination is not suitable for the type of detection. Maybe the contrast is rather low. You hope to use illumination to enhance the contrast? $\endgroup$ Aug 20, 2014 at 8:56
  • $\begingroup$ Yes, you should try to solve problems on the physical setup (imaging, illumination, mirrors etc.) as much as possible. Afterwards the code will be a breeze. For the solution, you need to use a homogeneous light setup to obtain a standardized image set to be able to use one solution for "all". Can you provide some images from the dataset ? $\endgroup$
    – Tyathalae
    Aug 20, 2014 at 9:34
  • $\begingroup$ I was told that the images are captured by the camera from one mobile phone. What you said is hard to do. I guess you are from machine vision field. For an machine vision application, more attention is paid to physical setup and so simple algorithms can solve the task. For my application, an complex algorithm is needed to extract the three curves. $\endgroup$ Aug 20, 2014 at 14:11
  • $\begingroup$ In medical image segmentation application, the contrast is low. Maybe some algorithms can be borrowed from medical image processing field. $\endgroup$ Aug 21, 2014 at 10:02


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.